DAC gets rolling: We hear from Gary Smith and learn about recent development in beyond CMOS technology.
It requires a certain dedication to attend technical DAC sessions on a Sunday morning, but full day workshops start before 9:00am for those dedicated to hearing about the latest work being conducted in academia and the research arm of industry. These are highly technical sessions that target academics and those serious about keeping a pulse on up and coming technologies.
One such workshop was titled “Design Automation for Beyond-CMOS Technologies.” Interestingly while many academic papers talk about the upcoming technologies such as spin logic, new transistors or 3D stacking, the keynote presented a set of challenges that have to be overcome before many of these technologies will find their way into commercial devices. Philip Wong, Professor of Electrical Engineering at Stanford University, identified several new technologies needed including a new switch, a new memory, new interconnect and new techniques for dealing with heat extraction from monolithic 3D stacks.
His talk identified a future that would consume 1/1000 of the energy profiles of current devices, meaning that the power of Watson could be available in a hand-held device, but it only comes about by using all of the new devices and changing the underlying architecture of the devices so that memory and logic are brought much closer together.
He admitted that the full solution may take 10 or 20 years to come to fruition even though parts of it are ready to start being integrated into devices today. Wong also pointed out some of the manufacturing challenges such as require low temperature manufacturing so that processing higher level of the stack do not damage the lower levels.
Following the keynote, several technical sessions talked about many of the individual pieces of the puzzle, presented as slides full of complex formulas, tables and diagrams. Summaries of the presentation will be available in the coming weeks.
One interesting concept being discussed considers that if the entirety of a device is constructed using non-volatile storage, then it changes the way we think about power. Consider a device where computation can proceed whenever power is available, potentially from scavenging, stop whenever sufficient power is not available and then continue uninterrupted when power is again restored.
Other talks concentrated on the new architectures that would be coupled with new device types, again showing that we are moving into an era when architecture and design are becoming as important as device advancements.
I was encouraged that the researchers were talking about real issues that would limit adoption of their work instead of presenting overly rosy views. Part of this stems from the fact that it takes multiple groups in industry and academia to make the necessary advancement. This more practical approach to problem solving makes it increasingly likely that these technologies will find their way into production in the future, and increased cooperation between the researchers will help find problems earlier in their work avoiding some of the unfruitful efforts associated with having overlook an important piece of the puzzle.
The early evening of DAC starts in the same way that it has for many years when Gary Smith presents his view of the industry and its most pressing problems. But this year is not to be the same as we are used to. Gary no longer had his white jacket, but did have orange socks and Harry Foster was sporting green and yellow check socks. Clearly fashion has changed for this year.
Gary’s principle message this year is that EDA is showing slow growth ($9B by 2019) assuming no downturns, and even when you add IP into the mix ($12B in 2019), it will not be enough to catch the interest of Wall Street. Wall Street wants to see steady growth with no surprises on either side.
IP has not fully shaken out yet, and there are additional pressures on royalties. IP ages quickly says Gary and what used to have value, is no longer going to bring in much money. To make any decent money, you having to be making platform IPs these days, and ARM is dominant in this market.
Because of this, EDA has begun looking beyond its borders into neighboring domains. This is perhaps not a new trend since Mentor has been in embedded software for almost 15 years, Synopsys into optical and application software markets and Mentor and Synopsys are also making forays into automotive, mechanical and other peripheral markets.
Gary thinks that mechanical is a fairly obvious choice, even though he pointed out that EDA is catching up and passing mechanical companies in terms of size. Other areas that should be considered are biomedical and chemical although the synergies have not been fully explored.
Gary made no mention of Cadence as a player in new markets, nor mentioned the emerging interest in security technologies that create opportunities in both hardware and software.
Are any of these a wise choice, especially when mechanical companies have shown interest in encroaching on EDA? Gary pointed out that EDA has been strong in algorithmic development although we are now losing some of these to financial and other Internet companies.
Gary actually managed to finish early although people stayed to ask questions instead of moving over to the packed reception area next door, whose volume increased steadily.